"Random" AI civs don't really seem random

Been playing binge hours G&K since I got it couple months ago. Had about 900 hrs vanilla and IIRC when I setup a game with AI opponents set to random, I got just that. Fairly random shuffle of original and DLC civs.

Fast forward 200+ hrs with G&K and it seems random is totally rigged. I always get G&K civs. With 34 civs, and me playing predominantly 6, 8, or 12 AI opponent setups, there should have been at least a good mix of games where I drew all non G&K opponents.

But in every game, maybe 25-30 games total played with another dozen or so messed around with briefly before restarting, I always draw Carthage, Ethiopia, etc. In fact, I can't recall a single game where either Carthage, Ethiopia, or Celts was not in it. Most frequently 2 of the 3.

Am I just on some incredible coincidence streak of random odds or does it seem G&K is hard coded to give you at least a couple G&K civs in every random game?

And, related to that, that you get at least 1 strong religious AI civ to prophet and missionary spam you. I've not had a single random game where Carthage, Ethiopia or Celts was not in it.

Now I feel sad, I never have Ethiopia in my games...

I think this is just a normal human misconception. I have played a of G&K games (minimum every civ once on at least standard sized maps), and while I sure had some "streaks" like Arabia and England appearing in at least every game for like 5-6 games, this is NORMAL for TRUE RANDOM. Remember, when humans are asked to create a random list where items can repeat (like a number code), they almost never repeat numbers or create a series (123, 555). In true random, those combinations are BOUND to happen sometimes, and because of the notoriety of the human brain to recognize patterns, humans instinctively feel something is "rigged" in a TRUE RANDOM shuffle. With true random I mean a list where Items can both repeat and are not divided in subgroups. And with my G&K experience, there is no preference for G&K civs, or any other. Heck, there is no preference for civs adapted to map type, which can give you Polynesia as opponent on Highland maps

Well I only got Germany as my CiV through random selection only a couple of days back after all these months. And shuffle seems to have a certain disposition for continents and Aggressive civs. But I am sure its just my luck.

Anecdotal evidence.

25-30 games is not a reasonable test.

The fact that you said 25-30 also says you kept not even a reasonable record to compare.

Random is random.

What proportion is the number of vanilla to G&K civs and in what proportion do you get vanilla and G&K civs in on average in a game?

That's only the tip of the myriad issues that need addressing before making a true statement

Come to think of it, I have never seen or played Ramesses or Augustus Caesar since vanilla...

Fluffball said:

Anecdotal evidence.

25-30 games is not a reasonable test.

The fact that you said 25-30 also says you kept not even a reasonable record to compare.

Random is random.

What proportion is the number of vanilla to G&K civs and in what proportion do you get vanilla and G&K civs in on average in a game?

That's only the tip of the myriad issues that need addressing before making a true statement

Click to expand...

Wow, what a snarky answer. Especially when so wrong.

Ok, didnt think I had to provide exact minutiae to be considered legitimate but all right.

I have EXACTLY 25 completed games with G&K because I can see them in my hall of fame, which for some reason got reset to zero when I installed G&K. So pretty easy to see and know all 25 games in my HoF are complete G&K games.

I THINK but don't have exact count with 100% assurance that I played approximately 5 games complete enough where I met all civs, but did not compete the game.

So, I have a sample size of minimum 25, with most likely 30.

There are 34 total civs with 9 of them introduced with G&K. That means when playing 25 sequential games, since sequential series means "dice have no memory", there is equal per game chance that for each first civ slot, there is a 9/34 or 26.5% chance for it to be populated with a G&K civ, and a 73.5% chance for it to be populated with non-GK civ.

However, per slot after the first, the odds of it being a non-GK civ increases while GK civ odds decrease, per slot.

Without going into the boring math unless that too is somehow a sufficient pretense for a snarky troll reply, the odds in 25 sequential games with at minimum 6 civ slots, with majority being 12 civ games, that ALL 25 samples will have Ethiopia, Celts, and/or Carthage in EVERY single game is astoundingly low.

Yes, it definitely can happen. And I'm open to the conclusion it could simply be simple chance. But I'm also open to possible being a preferential setting for G&K. What it definitely is not is the stupidly snarky AND wrong reply that a sample size of 25 is too small and because I have a range my point must not be valid for lack of record keeping. Any sample size that concludes the odds of that event occurring is less than a fraction of a fraction of 1% is in my opinion sufficient.

Sure, even then less than 1% chance events occur. Heck, for lotteries that would be great odds. But it would also be reasonable to question the event. Trolls serve no purpose on this forum other than to discourage new members such as myself from participating and hopefully contributing.

Well, had I not been playing G&K quite a bit lately, I would've said you experienced a mere coincidence - albeit a strange one. But I too have noticed the same EXACT thing you have. I have not had a single game without one of the three civs you mentioned. In fact, I've had games with ALL THREE in them, and I've had Carthage in the last 3 games I've played, with the Celts in the last 2.

Certainly interesting - not debilitating, per say - I get to study these civs quite well, haha. I wish I knew why this was the case, however.

I'm no math wiz, but if you have 12 AI civs in your game and if you divide your 33 possible opponents in 11 groups of 3, then I think the chance of getting Ethiopia OR Celts OR Carthage in 25 consecutive games is not that small.

And, to add my experience, I only have those 3 civs in my games comparatively rarely, and certainly not in every game.

But I have a different question: depending on the civ you choose, is there a chance for the "random" to be "random between flavors", where flavors are warmongers, ICS, wonder builders, etc? Although I doubt it, to tell you the truth.

From the programmer's point of view, I can tell you that it is SO MUCH simpler just to use a random generator, than use mess-ups, that I would be highly surprised if this is not what they did.

playshogi said:

I'm no math wiz, but if you have 12 AI civs in your game and if you divide your 33 possible opponents in 11 groups of 3, then I think the chance of getting Ethiopia OR Celts OR Carthage in 25 consecutive games is not that small.

Click to expand...

It gets bit more complicated if we want to be exact but rounded up rough math is in any one game where at least 3 civ slots are set to random, and assuming a true random generator, and no civ being allowed to be picked twice in same game -

a) with all 34 civs installed, there is a 1/34 or 2.9% chance with random selection of the first civ slot that it will be Ethiopia, a 2.9% chance for Celts, and 2.9% for Carthage.

b) technically, each slot thereafter has slightly adjusted odds because for example with the second civ slot, there would now only be 33 civs left to choose from. So a 3.0% chance it will either be Ethiopia, Celts, or Carthage

c) with the third slot it rises to 3.1% roughly.

d) since in a single event, the odds are cumulative, there would be a 2.9% + 3.0% + 3.1% chance, or roughly 9.0% chance that Ethiopia OR Celts OR Carthage will be present in any one game as long as there are a minimum of 3 civ slots selected for that game.

In that first sequential game, that's 9% for one, roughly 6% for any two of those civs present. I agree, not high, but not really low either.

However, the odds that you will have Ethiopia OR Celts OR Carthage in games #1 AND 2 is reduced to 0.08%, let's call it close to 1% for ease.

The odds over 25 sequential games to replicate this same condition assuming a truly random generator is a fraction of a fraction of 1%. Doesn't mean it can't happen, and by lottery odds it would be great. I remain open to the idea it could just be simple coincidence. With so many civ players, so many games, gotta happen to somebody. But I also remain open to G&K wanting to give you the "religious" xp with some hidden preference too.

This is a problem of combinatorial mathematics and this was a good opportunity for me to review it. If you are playing a huge game, you have 11 AI. You are playing 1 so that leaves 33 AI to choose from. This is like the casino game of keno where the casino draws 20 numbers from 80. In this case the computer is drawing 11 from 33. What are the odds that your 3 will not be drawn? There are 5456 combinations of 3 (33*32*31)/6.
You want all 3 to NOT be among the 11. There 1540 combinations of 3 from the 22 not drawn (22*21*20)/6. That means there is only 28% chance that 1 of the 3 won't be in the game. So, for 25 consecutive HUGE games for 1 of the 3 to be in it is .72 to the 25th power, which works out to be 1 chance in 3703. Which is small, but not millions to one, either.

Oops, you're right. I forgot to subtract myself from the civ list.

Either way pretty close to what I got. Pretty small odds but not impossible. Really good by lotto odds standards, think ill go buy a powerball ticket

playshogi said:

...which works out to be 1 chance in 3703. Which is small, but not millions to one, either.

Click to expand...

It still comes down to confirmation bias! Those are the odds when the 3 Civs are picked before running the 25 games. Jacozilla picked the three after he noticed a streak. The math escapes me, but maybe you are up for it: What are the odds that none of any three Civs won’t appear in all 25 games? I bet that is closer to 1 in 10^5.

beetle said:

It still comes down to confirmation bias! Those are the odds when the 3 Civs are picked before running the 25 games. Jacozilla picked the three after he noticed a streak. The math escapes me, but maybe you are up for it: What are the odds that none of any three Civs won’t appear in all 25 games? I bet that is closer to 1 in 10^5.

Click to expand...

Yes, it's important to note that the pattern was recognized after the fact. Presumably, the OP only finds Carthage, Ethiopia, and the Celts significant because they are G&K civs, and he noticed a pattern. I imagine had it been 3 other G&K civs, he would have made the same observation.

So while the odds of running into at least one of those specific civs in 25 consecutive games may have been 1 in ~3703, the odds of that pattern (in 25 consecutive games, having 3 G&K civs that appeared at least once) is notably higher. There are 11 G&K civs, and we are choosing 3. That is 165 combinations. So that's 165/3703 or 4.46%.

Edit: Note there are some overlapping combinations in the 165 * 1/3703 math, so it is less than 4.46% but still a reasonably frequent occurrence when you are talking about 1000s of people playing the game.

As other people have said, human misconception/low sample size. I always seem to get vanilla civs + one G+K or DLC civ. Someone else will never be able to get shot of Augustus. Another person will always get a completely even split of vanilla/G+K. For you to be able to safely conclude that 'random' is not truly random, you'd want to accurately record hundreds of games.

EDIT: Actually, I have a stats calculator to hand, let's go for it. I'll submit my attempt.

Jesus... So much math, can we just say:
Let Y represent Civ in question, and X represent Civs installed And C represent Map size.

Civ Slots. C = Standard/8 Civs
A civ being picked at random, the odds, are, the formula anyways, Y = 1/X-7.

For those who feel like reading it is: Civ in question has a 1 out of how many civs are installed -7 chance being picked of on a standard map.

Really really random.

As in, setting everything to random, and I get this weird shaped island. On which I'm alone as Russia, way up in the north. It's archipelago, standard size. So, off to meet everyone else and...

...there's Denmark.
...and Polynesia.
...and Askia.
...and the Ottomans.

Well, there were the odd ones out, like the Huns, Alex and the Swedes. But funny random. I didn't get England unfortunately so there goes another chance to send them my love from Moscow.

Heh Heh, stat maths will always get wheeled out at some point in such a debate.
I don't believe that it is entirely random, I think the answer suggesting flavours matches my experiences; and this would make a lot of sense in terms of 'playability'.

Confirmation bias: you'll only recall the games/instances where your suspicion (of only having G&K civs) was confirmed, easily forgetting the other nrmal games, making it seems disproportianal